METHODS & TECHNIQUES

A sample preparation workflow for adipose tissue shotgunproteomics and proteogenomicsJane I. Khudyakov1,2,*, Jared S. Deyarmin1, Ryan M. Hekman1, Laura Pujade Busqueta1, Rasool Maan1,Melony J. Mody1, Reeti Banerjee1, Daniel E. Crocker3 and Cory D. Champagne2

ABSTRACTAnimals with large adipose stores, such as marine mammals, mayprovide insights into the evolution and function of this multifunctionaltissue in health and disease. In the absence of sequenced genomes,molecular information can be rapidly obtained by proteomics andtranscriptomics, but their application to adipose tissue is hindered bylow nucleic acid and protein yields. We sequenced and comparedproteomes isolated from the blubber of four elephant seals usingphenol and guanidine thiocyanate (Qiazol) or detergent (sodiumdeoxycholate) buffer. Qiazol recovered more subcellular proteinssuch as metabolic enzymes, in addition to extracting RNA, facilitatingproteogenomic analyses of small lipid-rich tissue biopsies. We alsocompared proteomics data analysis platforms and found that de novopeptide sequencing improved protein identification sensitivitycompared to database search alone. We report sample preparationand data analysis workflows for proteogenomics and a proteome ofelephant seal blubber containing 2678 proteins, including many ofinterest for further functional studies.

This article has an associated First Person interview with the firstauthor of the paper.

KEY WORDS: Adipose, Blubber, Proteomics

INTRODUCTIONAdipose is a complex organ that participates in energy storage,thermogenesis, immunity and regulation of metabolic homeostasis.Specialized fat deposits arose early in the metazoan lineage andhave supported evolutionary adaptations such as migration,hibernation and lactation, among others (Birsoy et al., 2013;Pond, 2017). While adipose tissue has been studied extensively inhumans and laboratory animals due to the emergence of a globalobesity epidemic, non-model organisms can provide fundamentalinformation on metabolic adaptations in animals and potentiallynovel insights into mechanisms by which adipose function isdysregulated in disease (Grabek et al., 2015; Houser et al., 2013).Such insights can be rapidly obtained via non-targeted approaches

such as transcriptomics and proteomics in the absence of availablegenomes (Nesvizhskii, 2014).

With some of the largest subcutaneous adipose stores (modifiedas blubber) in the animal kingdom, marine mammals hold valuableinformation about rapid fat accrual and loss, lipid-based metabolismand the physiological effects of lipophilic pollutants (Bossart, 2011;Houser et al., 2013). Several recent studies have used omicstechnologies to profile gene and protein expression in blubber(Brown et al., 2017; Kershaw et al., 2018; Khudyakov et al., 2017;Van Dolah et al., 2015). However, widespread application of theseapproaches to marine mammal systems is hindered by the technicalchallenge of obtaining sufficient quantities of nucleic acids andproteins from small biopsies of lipid-rich tissues with low nuclearand cytoplasmic content. Indeed, most proteomics studies of marinemammals to date have used tissue matrices other than blubber(Neely et al., 2015a, b; Sobolesky et al., 2016).

Nucleic acids and proteins are commonly isolated from cells andtissues using separate pipelines, with phenol and guanidinethiocyanate solutions (e.g. Trizol®) used for RNA and DNAextraction and detergents (e.g. sodium dodecyl sulfate, SDS) forprotein isolation (Feist and Hummon, 2015; Zhang et al., 2013). TheTrizol® reagent was originally developed for simultaneous extractionof DNA, RNA, and proteins (Chomczynski, 1993), but itseffectiveness for RNA and protein extraction from adipose, andproteome completeness and compatibility with tandem massspectrometry (MS/MS), have not yet been described. In this study,we compared two lysis buffers for shotgun proteome sequencing ofadipose tissue: (1) a detergent buffer containing sodiumdeoxycholate(SDC method), and (2) Qiazol®, a solution similar to Trizol® thatwas developed for nucleic acid isolation from lipid-rich samples (QIAmethod). We used blubber samples collected from four juvenilenorthern elephant seals (Mirounga angustirostris), a fasting-adaptedmarine mammal study system frequently used in comparativemetabolic physiology studies, including analyses of cellularresponses to stress and fasting (Khudyakov et al., 2017; Martinezet al., 2018). Marine mammal blubber is vertically stratified by fattyacid composition and function; the outer layer plays a role inthermoregulation while the inner layer is more metabolically active(Strandberg et al., 2008).We used the outer half of blubber biopsies inthis study, as this layer is typically sampled by remote biopsy dartfrom many marine mammals (Hunt et al., 2013).

We show that a larger number of unique proteins, includingthose involved in metabolism and protein translation, can beidentified in samples lysed using QIA with the added benefitof RNA isolation, and that de novo peptide sequencing(PEAKS Studio) combined with database search increasessensitivity of protein identification compared with databasesearch alone (SEQUEST in Proteome Discoverer). We report thefirst elephant seal outer blubber layer proteome containing anumber of metabolic enzymes and adipokines of interest toReceived 19 June 2018; Accepted 5 October 2018

1Department of Biological Sciences, University of the Pacific, Stockton, CA, USA.2Conservation and Biological Research Program, National Marine MammalFoundation, San Diego, CA, USA. 3Department of Biology, Sonoma StateUniversity, Rohnert Park, CA, USA.

*Author for correspondence ( jkhudyakov@pacific.edu)

J.I.K., 0000-0001-7038-102X; D.E.C., 0000-0002-7940-8011; C.D.C., 0000-0003-3938-4462

This is an Open Access article distributed under the terms of the Creative Commons AttributionLicense (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use,distribution and reproduction in any medium provided that the original work is properly attributed.

1

© 2018. Published by The Company of Biologists Ltd | Biology Open (2018) 7, bio036731. doi:10.1242/bio.036731

BiologyOpen

by guest on July 12, 2020http://bio.biologists.org/Downloaded from

comparative physiologists and provide an optimized protocol forproteogenomics studies of adipose tissue.

RESULTS AND DISCUSSIONProtein yieldThe study was conducted using four biological replicates of blubbertissue. Proteins were isolated from one half of each sample using theSDC method and RNA and proteins were isolated from the otherhalf using the QIA method. SDC yielded 1.85-fold more proteinthan QIA (paired t-test, t=4.48, P<0.05). Mean protein yields permg wet mass tissue for SDC and QIA were 13.7 µg/mg (s.d.=1.8)and 7.4 µg/mg (s.d.=2.2), respectively. Lower protein yields havebeen reported for Trizol® compared with detergent-based buffers(Yamaguchi et al., 2013). However, QIA also recovered mean4.6 µg (s.d.=4.2) of total RNA per sample. Variability in RNAyields between samples (range: 1.78–10.89 µg) could be due todecreased efficiency of homogenization and silica column-basedRNA purification with higher tissue inputs. RNA isolated by QIAfrom blubber samples had high purity and integrity. The mean 260/280 and 260/230 ratios were 1.99 (s.d.=0.07) and 1.75 (s.d.=0.25),respectively. The mean RNA integrity number (RIN) was 7.67(s.d.=0.39) and mean rRNA ratio was 0.81 (s.d.=0.12) (Fig. S1).RNA integrity was above the threshold (RIN 7) commonlyrecommended for RNA sequencing (Gallego Romero et al., 2014).

Protein identificationAfter processing using standard methods [Fig. 1; Bodzon-Kulakowska et al. (2007)], protein samples were analyzed byHPLC-MS/MS, producing mean 26,621 (s.d.=1031) MS/MS spectrafor SDC samples and mean 25,712 (s.d.=1374) MS/MS spectra forQIA samples. We first performed peptide spectrum matching and aSwissProt database search using SEQUEST in Proteome Discoverer.SEQUEST identified 13.3% and 16.1% of all MS/MS spectra fromSDC and QIA samples, respectively, which is within the range of 10–30% reported in the literature (Houel et al., 2010). QIA samples had1.17-fold more peptide spectrum matches (PSMs; F1,10=100.45,P<0.0001), 1.13-fold more peptide groups (F1,10=67.01, P<0.0001),1.21-fold more protein groups (F1,10=59.09, P<0.0001), and 1.25-fold more unique proteins (with two or more unique peptide hits;F1,10=24.54,P<0.001) than SDC samples (Fig. 2A). Therefore, whiletotal protein yields were lower with QIA, this method produced moreidentified peptides and proteins than SDC. We then repeated theSEQUEST searchwith a custom database of a translated elephant sealblubber transcriptome (Khudyakov et al., 2017). QIA-isolatedsamples had 6048 PSMs (s.d.=373), 4620 peptide groups(s.d.=383), 974 protein groups (s.d.=53) and 647 unique proteins(s.d.=45), while the SDC-isolated samples had 5473 PSMs(s.d.=172), 4348 peptide groups (s.d.=174), 865 protein groups(s.d.=46) and 564 unique proteins (s.d.=31). Numbers of PSMs andproteins were significantly different between methods (PSMs:F1,3=17.16, P<0.05; protein groups: F1,3=18.52, P<0.05; uniqueproteins: F1,3=16.33, P<0.05). Therefore, we identified hundreds ofproteins predicted from the elephant seal transcriptome by MS/MS,validating our transcriptome and proteome methods and providing aworkflow for proteogenomics. However, since the translatedtranscriptome was annotated by BLASTP using the same database,we focused subsequent functional analyses on proteins identifieddirectly by SEQUEST SwissProt database search.

Functional annotation of proteinsTo facilitate functional analyses, we repeated SEQUEST SwissProtdatabase search with biological replicates concatenated as ‘fractions’

into one pooled sample for each QIA and SDC. We identified 976and 701 proteins from pooled QIA and SDC samples, respectively(2.5-fold more unique proteins with QIA than SDC; Fig. 2B). Thetop 5 (by number of proteins) KEGG categories overrepresented inboth datasets were focal adhesion, PI3K-Akt signaling pathway,biosynthesis of antibiotics (which includes many enzymes involvedin lipid metabolism), ECM-receptor interaction, and carbonmetabolism (Fig. 3A). KEGG categories enriched uniquely in theQIA dataset were ribosome, pyruvate metabolism, pentosephosphate pathway, valine, leucine and isoleucine degradation,while those unique to SDC were platelet activation and small celllung cancer. Therefore, the QIA method recovered more proteinsinvolved in metabolism and protein synthesis than SDC.

We identified 37 and 41 GO biological process (BP) categoriesenriched in the QIA and SDC datasets, respectively (Fig. 3B). Thetop overrepresented categories in both datasets were associated withcell-matrix interactions and protein folding. BP categories unique tothe QIA dataset included translation, mRNA splicing, rRNAprocessing, gene expression, pentose-phosphate shunt andresponse to calcium ion. BP categories that were enriched only inthe SDC dataset included osteoblast differentiation, epithelial celldifferentiation, muscle contraction, MAPK cascade, tumor necrosisfactor-mediated signaling pathway, cellular response totransforming growth factor beta stimulus, regulation of cellmigration, nucleosome assembly and blood coagulation. Whileboth methods recovered proteins involved in cell–cell and cell–matrix adhesion, QIA isolated additional proteins were involved inmRNA processing and protein synthesis, while SDC recoveredmore proteins involved in cell signaling and differentiation.

There were 22 and 16 GO molecular function (MF) categoriesenriched in the QIA and SDC datasets, respectively (Fig. 3C). TopMF categories for each method were protein and RNA binding andcadherin-mediated cell–cell adhesion. MF categories enriched onlyin the QIA dataset were nucleotide binding, structural constituent ofribosome, protein complex binding, heparin binding and chaperonebinding. MF categories unique to the SDC dataset wereoxidoreductase activity and platelet-derived growth factor binding.Therefore, while representation ofmolecular functions was similar inboth protein datasets, some proteins with oxidoreductase activity ofinterest to marine mammal physiology (e.g. fatty acid synthase,peroxiredoxin, alcohol dehydrogenase) were not isolated by QIA.

We identified 50 and 48 GO cellular component (CC) categoriesthat were enriched in the QIA and SDC datasets, respectively(Fig. 3D). Top CC categories for both methods were extracellularexosome, cytoplasm/cytosol, membrane, and extracellular space/matrix. CC categories enriched only in the QIA dataset includedperinuclear region of cytoplasm, mitochondrial matrix, ribosome,extrinsic component of membrane, spliceosomal complex andnuclear matrix. CC categories enriched only in the SDC datasetwere actin cytoskeleton, protein complex, cell–cell junction,lysosomal lumen, sarcolemma, lipid particle, smooth endoplasmicreticulum and lamin filament. The large number of extracellularproteins recovered by both methods is consistent with thick basallamina and abundant connective tissue proteins characteristic ofadipose tissue (Mariman and Wang, 2010). The abundance ofextracellular vesicle (EV)-related proteins recovered by bothmethods is interesting due to their potential role in regulation ofmetabolism and immunity in adipose tissue (Gao et al., 2017).However, the QIA method isolated more subcellular proteins thanSDC, including those associated with the nucleus andmitochondria,suggesting that it may be more efficient at solubilizing intracellularmembranes than detergent.

2

METHODS & TECHNIQUES Biology Open (2018) 7, bio036731. doi:10.1242/bio.036731

BiologyOpen

by guest on July 12, 2020http://bio.biologists.org/Downloaded from

Fig. 1. Sample preparation workflows used in the study. Elephant seal blubber samples were divided in half and each was lysed by bead beating witheither Qiazol® (QIA method) or buffer containing sodium deoxycholate detergent (SDC method). After protein precipitation from tissue lysates, samples weretreated identically (grey boxes). MS/MS data was analyzed by SEQUEST database search in Proteome Discoverer or by de novo sequencing and PEAKSDB database search in PEAKS Studio.

3

METHODS & TECHNIQUES Biology Open (2018) 7, bio036731. doi:10.1242/bio.036731

BiologyOpen

by guest on July 12, 2020http://bio.biologists.org/Downloaded from

De novo peptide sequencingWe performed de novo peptide sequencing combined with PEAKSDB SwissProt database search (Zhang et al., 2012) to evaluate

whether this approach would increase sensitivity of peptide/proteinidentification. PEAKS produced 4534 (s.d.=567) and 4443(s.d.=671) de novo-only spectra for samples prepared using SDC

Fig. 2. Comparison of proteins identified using two sample preparation and two data analysis workflows. (A) Numbers of peptide spectrum matches(PSMs) and peptides, protein groups, and unique proteins (with two or more unique peptide hits) identified by SEQUEST or PEAKS in samples preparedusing SDC and QIA. SwissProt (2/13/2018) database was used for searches and only hits with false discovery rates (FDR) <1% were retained. Asterisksdenote significant differences between sample preparation methods and software platforms: ***P<0.0001, **P<0.001. Sets of identified proteins werecompared between (B) SDC and QIA methods and SEQUEST database search, (C) SEQUEST and PEAKS protein identification methods for QIA samples,and (D) between each sample preparation method and software platform.

4

METHODS & TECHNIQUES Biology Open (2018) 7, bio036731. doi:10.1242/bio.036731

BiologyOpen

by guest on July 12, 2020http://bio.biologists.org/Downloaded from

andQIA, respectively. For both sample preparationmethods, PEAKShad approximately 1.10-fold more PSMs (F1,10=44.27, P<0.0001)and identified 1.27-fold more peptides (F1,10=242.24, P<0.0001) and 2.24-fold more unique proteins (F1,10=173.73,P<0.0001) than SEQUEST (Fig. 2A). However, the number ofprotein ‘groups’ identified by PEAKS was approximately 1.95-foldlower than SEQUEST (F1,10=572.21, P<0.0001) due to differences

in protein isoform clustering approaches (Paulo, 2013). Therefore,combined de novo peptide sequencing and PEAKS DB databasesearch had greater sensitivity than a SEQUEST database searchalone, as previously reported (Zhang et al., 2012).

Lastly, we compared the sets of unique proteins identified bySEQUEST and PEAKS using the SwissProt database in the pooledQIA samples. The two software platforms identified a common set

Fig. 3. Comparison of functional categories enriched in protein datasets obtained using two sample preparation workflows. Top (A) KEGGpathways and (B) gene ontology (GO) biological process, (C) molecular function, and (D) cellular component categories that were significantlyoverrepresented (adjusted P<0.05) in the elephant seal blubber proteome isolated using either QIA (yellow) or SDC (blue) methods, relative to the entirehuman genome.

5

METHODS & TECHNIQUES Biology Open (2018) 7, bio036731. doi:10.1242/bio.036731

BiologyOpen

by guest on July 12, 2020http://bio.biologists.org/Downloaded from

of 595 proteins; an additional 1258 were identified only by PEAKSand 381 were identified only by SEQUEST (Fig. 2C). Differentprotein identification approaches and database search algorithms areknown to produce different sets of proteins from the same massspectra, even with identical search parameters (Paulo, 2013;Russeth et al., 2006). Integrating results from multiple searchengines can increase the number of identified proteins and validatethose that are commonly identified by different algorithms, andseveral bioinformatics tools have been developed for this purpose(Paulo, 2013).Overall, we identified overlapping, but distinct sets of proteins

from blubber samples using two different sample preparationmethods and two different MS/MS data analysis platforms(Fig. 2D). Proteins of interest to the metabolic and comparativephysiology communities identified in the study include fatty acidtransporters, lipid droplet proteins, and lipid metabolism enzymes(Table S3). In total, we identified 2678 proteins from the outerblubber of northern elephant seals, of which 286 were common toall four pipelines used in the study. Differences in numbers andtypes of proteins isolated by the QIA and SDC methods may beattributed to different membrane solubility efficiencies – reagentssuch as Trizol® may be more efficient than detergents at removinglipids and carbohydrates to liberate proteins (Kirkland et al., 2006).However, some have suggested that tissue lysis in phenol andguanidine thiocyanate may also lead to loss of highly hydrophobicproteins, an important consideration for adipose proteomics (Buttet al., 2007; Kirkland et al., 2006). This may be improved byoptimizing solubilization conditions for proteins precipitated afterTrizol® or Qiazol® extraction (Kopec et al., 2017).

CONCLUSIONSWe found that blubber tissue lysis in Qiazol® increased the totalnumber of identified proteins and enabled simultaneous isolation ofhigh-quality RNA from the same tissue sample – significantadvantages for researchers working with small quantities of tissueand for those interested in proteogenomics. Moreover, QIArecovered more subcellular proteins, including proteins involvedin metabolism and protein synthesis, than SDC. We also found thatsensitivity of protein identification in a non-model organism couldbe significantly improved using PEAKS de novo peptidesequencing in combination with a database search. Proteinsidentified in this study may be of interest for functional studies ofmarine mammals and other species in which large adipose storesplay key roles in physiology.

MATERIALS AND METHODSMaterialsAll chemicals were proteomics grade and purchased from VWR LifeScience/Amresco (USA) or Thermo Fisher Scientific (USA), unlessotherwise indicated.

Sample collectionAll animal handling procedures were approved by University of the Pacificand Sonoma State University Institutional Animal Care and UseCommittees and conducted under National Oceanic and AtmosphericAdministration Fisheries Permit No. 19108. Four juvenile (∼0.8-year old)female northern elephant seals (M. angustirostris) were sampled at AñoNuevo State Reserve (San Mateo County, CA, USA) in October 2017.Animals were chemically immobilized as previously described (Khudyakovet al., 2017). Blubber samples were collected from the posterior flank of theanimal using a sterile 6.0 mm diameter biopsy punch (Miltex, USA), blottedon sterile gauze to remove blood, and separated into two halves: an inner(closest to muscle) blubber half and an outer (closest to skin) blubber half.

Samples were placed into plastic cryogenic vials (Corning, USA), flash-frozen in liquid nitrogen, stored on dry ice, and transferred to a −80°Cfreezer upon return to the laboratory.

Sample preparationOnly the outer blubber half of each biopsy was used for this study (innerblubber was used for a separate study). Each outer blubber sample wasweighed and minced into small pieces, which were randomly divided intotwo portions of approximately 100 mg each. One portion of each samplewas used for protein extraction using the detergent method (mean wet mass97.5 mg, s.d.=8.6), while the other was used for RNA and protein extractionusing the Qiazol® method (mean wet mass 102.8 mg, s.d.=9.0).

Protein extraction using detergent (SDC method)The SDC method of protein extraction was adapted from a previouslypublished protocol (Pasing et al., 2017). Blubber was processed in twobatches of ∼50 mg, which were minced with a sterile scalpel on ice andadded to 500 µl SDC Lysis Buffer [1% w/v SDC, 8 M urea, 5 mMdithiothreitol (DTT) in 50 mM ammonium bicarbonate] in a Navy RINO®

bead tube (Next Advance Inc., USA). Samples were homogenized in theBullet Blender Storm® instrument (Next Advance Inc., USA) for two cyclesof 2 min each at power 10, with 1 min of cooling on ice between cycles.Homogenates were further disrupted by sonication for three cycles, 15 seach, at 4 watts using a hand-held sonicator (VirSonic 60, Virtis, USA) andcentrifuged to pellet insoluble cell debris and separate lipids. Tissuehomogenates were extracted from under the top lipid layer and transferred toclean tubes. To remove any remaining lipids, four volumes of methanol andone volume of chloroform were added to homogenate aliquots, mixed, andcentrifuged. The top layer containing lipids was removed, and four volumesof methanol were added to precipitate proteins. Protein pellets were air driedafter centrifugation.

Protein and RNA extraction using phenol-chloroform (QIAmethod)The QIA method of protein extraction was adapted from the Chomczynskiprotocol (Chomczynski, 1993). Approximately 100 mg of blubber wereminced with a sterile scalpel on ice and added to 500 µl of Qiazol® reagent(Qiagen, USA) in a Navy RINO® RNase-free bead tube (Next Advance,USA). Homogenization was conducted in the Bullet Blender Storm® (NextAdvance, USA) as described above. An additional 500 µl of Qiazol®reagent was added to each tube and incubated for 5 min at room temperaturewith occasional vortexing. Homogenates were further disrupted using asyringe and 21-gauge needle and centrifuged to pellet insoluble cell debrisand separate lipids. The homogenate was extracted from under the top lipidlayer and transferred to clean tubes with 200 µl chloroform, vortexed tomix, and centrifuged at maximum speed for 15 min at 4°C to separatephases. The aqueous layer containing RNA was transferred to a clean tubeand RNA purification was performed using RNeasy® Lipid Tissue MiniKit (Qiagen, USA) (Khudyakov et al., 2017). After aqueous phase andinterphase extraction, 300 µl of 100% ethanol was added to the organicphase and centrifuged to precipitate DNA. The supernatant was split intotwo microcentrifuge tubes and each was incubated for 10 min with 750 µlof 100% isopropanol and centrifuged again to pellet proteins. The pelletswere washed twice for 20 min with 1 ml of 0.3 M guanidine hydrochloridein 95% ethanol, once with 1 ml of 100% ethanol and air dried. If notprocessed the same day, protein pellets were kept at −80°C in guanidinewash solution.

Protein denaturation, digestion, and desaltingProteins isolated by SDC and QIA methods were treated identically afterprecipitation. Pellets were resuspended in 250 µl of Denaturing Buffer (1%w/v SDC, 8 M urea, 5 mM DTT in 50 mM ammonium bicarbonate) byhomogenization with a 21-gauge needle and continuous vortexing for 1 h atroom temperature. Protein samples were then incubated at 37°C for 1 h inDenaturing Buffer, followed by alkylation with 15 mM iodoacetamide inthe dark at room temperature for 30 min. Alkylation was quenched byaddition of DTT to 5 mM final concentration. Samples were diluted with

6

METHODS & TECHNIQUES Biology Open (2018) 7, bio036731. doi:10.1242/bio.036731

BiologyOpen

by guest on July 12, 2020http://bio.biologists.org/Downloaded from

50 mM ammonium bicarbonate to reduce urea concentration to <2 M andprotein concentration was estimated by bicinchoninic acid assay (BCAassay) as described below. In-solution trypsin digest was conducted for 14–16 h at 37°C using Trypsin Gold®, Mass Spectrometry Grade (Promega) at1:50 of µg enzyme to µg protein. Samples were acidified to pH<2 withtrifluoroacetic acid to precipitate detergent and desalted using Pierce C18Spin Columns (Thermo Fisher Scientific). To maximize protein retention,the flow-through after the addition of the sample to the column was passedback over the column twice. Proteins were eluted with 70% acetonitrile,diluted 1:1 with HPLC-grade water, lyophilized, and resuspended in 0.1%formic acid in LC/MS-grade water. Peptide concentration was estimated byBCA assay as described below.

BCA assayProtein and peptide concentration was estimated using Pierce BCA ProteinAssay Kit (Thermo Fisher Scientific). Samples were diluted 1:10 in 50 mMammonium bicarbonate and each sample was used in triplicate in the BCAAssay. The mean coefficient of variation (CV) for triplicates was 1.57%.Prism7 (GraphPad, USA) was used to fit the curve (third-order polynomialfit, r2=0.99) and extrapolate unknown sample concentrations.

RNA quantity and quality assessmentRNA yields were determined using RNA BR Assay on the Qubit 3.0fluorometer (Life Technologies). RNA quality was evaluated by NanoDropspectrophotometry and microcapillary gel electrophoresis (RNA 6000 Picoassay, Bioanalyzer 2100 instrument, Agilent, USA).

HPLC-MS/MSPeptide samples were diluted to 150 ng/µl in 0.1% formic acid in LC/MS-grade water and 5 μl were loop injected by a Dionex Ultimate 3000autosampler onto a reversed-phase trap column (Acclaim® PepMap® 100C18 LC column; 75 µm i.d.×2 cm, 3 µm particle size, 100 A pore size,Thermo Fisher Scientific), and eluted onto a reversed-phase analyticalcolumn (EASY-Spray® C18 LC column; 75 µm i.d.×15 cm, 100 A, ThermoFisher Scientific) held at 35°C. Solvents A and B were 0.1% formic acid inwater and in acetonitrile, respectively. Solvent B was used at the followingconcentrations: 2% for 5 min, 2–22% over 70 min, 22–38% over 25 min, 38–95% over 5 min, 95% for 5 min, return to 2% over 5 min, 2% for 25 min.Flow rates were held at 300 nl/min with each sequencing run set to 140 min.Mass spectrometry analysis was performed using Orbitrap Fusion® Tribrid®

mass spectrometer equipped with EASY-Spray® ion source (Thermo FisherScientific) operated in a data dependent acquisition (DDA) mode by Xcalibur4.0 software (Thermo Fisher Scientific). Briefly, full MS1 scans wereresolved by the orbitrap, and ions were selected for MS2 using DDA (chargestate: 2–7; intensity threshold: 25,000). These precursor ions were quadrupolefiltered and subsequently fragmented using stepped collision HCD at 28%±3collision energy. MS2 product ions were resolved by the orbitrap. Instrumentand data acquisition settings are presented in Table S1.

MS/MS data analysisMS/MS data was analyzed using Proteome Discoverer v2.2 (Thermo FisherScientific) and PEAKS Studio v8.5 (Bioinformatics Solutions Inc., USA).Peptide spectra were searched against the entire UniProt SwissProt database(downloaded on 2/13/2018) concatenated with a common contaminantdatabase (common Repository of Adventitious Proteins, cRAP, https://www.thegpm.org/crap/index.html) using (1) SEQUEST in ProteomeDiscoverer or(2) PEAKS DB after de novo sequencing. Search parameters are shown inTable S2. False discovery rate (FDR) was estimated by searching a reversedconcatenated database in Proteome Discoverer and by the ‘decoy-fusion’approach in PEAKS (Zhang et al., 2012). Results were filtered to retainpeptides and proteins with a false discovery rate (FDR) <1% and to removethose with hits to the contaminant database. Proteins were considered‘unique’ if they had two or more unique peptides that mapped to them.DAVID Bioinformatics Resources v6.8 (Huang da et al., 2009) server wasused to identify KEGG and GO categories that were overrepresented in theseal blubber proteome relative to the entire human genome (P<0.05, adjustedfor multiple comparisons using Benjamini correction).

Statistical analysesAll statistical analyses were conducted using JMP 13 (SAS Institute Inc.,USA). Protein yields were compared by paired t-test (two-tailed), assumingunequal variances. The numbers of PSMs, peptides, proteins, and proteingroups were compared using linear mixed models with method (SDC orQIA) and software (SEQUEST or PEAKS) as fixed effects and sample ID asa random effect.

AcknowledgementsThe authors are grateful to C. Vierra, G. Knudsen, D. Pappin and K. Rivera for adviceon methods development and data analysis. The authors thank D. Costa,P. Robinson and the park rangers at An o Nuevo State Park and UC Santa Cruz’sAn o Nuevo State Reserve for facilitating access to the animals and to E. Abdollahi,D. Houser, M. McCormley, A. Ngo, G. Sandhu and A. Stephan for assistance withsample collection.

Competing interestsThe authors declare no competing or financial interests.

Author contributionsConceptualization: J.I.K.; Methodology: J.I.K., J.S.D., R.M.H., L.P.B., R.M., M.J.M.,R.B.; Validation: J.I.K., J.S.D., R.M.H.; Formal analysis: J.I.K., J.S.D., L.P.B., R.M.,M.J.M., R.B.; Investigation: J.I.K., J.S.D., R.M.H., L.P.B., C.D.C., D.E.C.;Resources: J.I.K., R.M.H.; Data curation: J.I.K., J.S.D.; Writing - original draft: J.I.K.;Writing - review & editing: J.I.K., J.S.D., R.M.H., L.P.B., C.D.C., D.E.C.;Visualization: J.I.K., R.M., M.J.M., R.B.; Supervision: J.I.K.; Project administration:J.I.K.; Funding acquisition: J.I.K., C.D.C., D.E.C.

FundingThis work was supported by the Office of Naval Research [N00014-15-1-2773 toJ.I.K., C.D.C., and D.E.C.].

Data availabilityRaw spectra and peak lists were submitted to MassIVE proteomics repository,accession #MSV000082455, download from ftp://MSV000082455@massive.ucsd.edu. Database search results and de novo proteinsequences were submitted to Figshare: https://figshare.com/s/ad1bf8a2d6824e617193.

Supplementary informationSupplementary information available online athttp://bio.biologists.org/lookup/doi/10.1242/bio.036731.supplemental

ReferencesBirsoy, K., Festuccia, W. T. and Laplante, M. (2013). A comparative perspective

on lipid storage in animals. J. Cell Sci. 126, 1541.Bodzon-Kulakowska, A., Bierczynska-Krzysik, A., Dylag, T., Drabik, A., Suder,

P., Noga, M., Jarzebinska, J. and Silberring, J. (2007). Methods for samplespreparation in proteomic research. J. Chromatogr. B Analyt. Technol. Biomed. LifeSci. 849, 1-31.

Bossart, G. D. (2011). Marine mammals as sentinel species for oceans and humanhealth. Vet. Pathol. 48, 676-690.

Brown, T. M., Hammond, S. A., Behsaz, B., Veldhoen, N., Birol, I. and Helbing,C. C. (2017). De novo assembly of the ringed seal (Pusa hispida) blubbertranscriptome: a tool that enables identification of molecular health indicatorsassociated with PCB exposure. Aquat. Toxicol. 185, 48-57.

Butt, R. H., Pfeifer, T. A., Delaney, A., Grigliatti, T. A., Tetzlaff, W. G. andCoorssen, J. R. (2007). Enabling coupled quantitative genomics and proteomicsanalyses from rat spinal cord samples. Mol. Cell. Proteomics 6, 1574-1588.

Chomczynski, P. (1993). A reagent for the single-step simultaneous isolation ofRNA, DNA and proteins from cell and tissue samples. Biotechniques 532-4,536-537.

Feist, P. and Hummon, A. B. (2015). Proteomic challenges: sample preparationtechniques for microgram-quantity protein analysis from biological samples.Int. J. Mol. Sci. 16, 3537-3563.

Gallego Romero, I., Pai, A. A., Tung, J. and Gilad, Y. (2014). RNA-seq: impact ofRNA degradation on transcript quantification. BMC Biol. 12, 42-42.

Gao, X., Salomon, C. and Freeman, D. J. (2017). Extracellular vesicles fromadipose tissue—a potential role in obesity and type 2 diabetes? Front. Endocrinol.(Lausanne) 8, 202.

Grabek, K. R., Martin, S. L. and Hindle, A. G. (2015). Proteomics approaches shednew light on hibernation physiology. J. Comp. Physiol. B 185, 607-627.

Houel, S., Abernathy, R., Renganathan, K., Meyer-Arendt, K., Ahn, N. G. andOld, W. M. (2010). Quantifying the impact of chimera MS/MS spectra on peptideidentification in large scale proteomics studies. J. Proteome Res. 9, 4152-4160.

7

METHODS & TECHNIQUES Biology Open (2018) 7, bio036731. doi:10.1242/bio.036731

BiologyOpen

by guest on July 12, 2020http://bio.biologists.org/Downloaded from

Houser, D. S., Champagne, C. D. and Crocker, D. E. (2013). A non-traditionalmodel of themetabolic syndrome: the adaptive significance of insulin resistance infasting-adapted seals. Front.Endocrinol. 4, 164.

Huang da, W., Sherman, B. T. and Lempicki, R. A. (2009). Systematic andintegrative analysis of large gene lists using DAVID bioinformatics resources. Nat.Protoc. 4, 44-57.

Hunt, K. E., Moore, M. J., Rolland, R. M., Kellar, N. M., Hall, A. J., Kershaw, J.,Raverty, S. A., Davis, C. E., Yeates, L. C., Fauquier, D. A. et al. (2013).Overcoming the challenges of studying conservation physiology in largewhales: areview of available methods. Conserv. Physiol. 1, cot006.

Kershaw, J. L., Botting, C. H., Brownlow, A. and Hall, A. J. (2018). Not just fat:investigating the proteome of cetacean blubber tissue. Conserv. Physiol. 6,coy003.

Khudyakov, J. I., Champagne, C. D., Meneghetti, L. M. and Crocker, D. E.(2017). Blubber transcriptome response to acute stress axis activation involvestransient changes in adipogenesis and lipolysis in a fasting-adapted marinemammal. Sci. Rep. 7, 42110.

Kirkland, P. A., Busby, J., Stevens, S. and Maupin-Furlow, J. A. (2006). Trizol-based method for sample preparation and isoelectric focusing of halophilicproteins. Anal. Biochem. 351, 254-259.

Kopec, A. M., Rivera, P. D., Lacagnina, M. J., Hanamsagar, R. and Bilbo, S. D.(2017). Optimized solubilization of TRIzol-precipitated protein permits Westernblotting analysis to maximize data available from brain tissue. J. Neurosci.Methods 280, 64-76.

Mariman, E. C. M. andWang, P. (2010). Adipocyte extracellular matrix composition,dynamics and role in obesity. Cell. Mol. Life Sci. 67, 1277-1292.

Martinez, B., Khudyakov, J., Rutherford, K., Crocker, D. E., Gemmell, N. andOrtiz, R. M. (2018). Adipose transcriptome analysis provides novel insights intomolecular regulation of prolonged fasting in northern elephant seal pups. Physiol.Genomics 50, 495-503.

Neely, B. A., Ferrante, J. A., Chaves, J. M., Soper, J. L., Almeida, J. S., Arthur,J. M., Gulland, F. M. D. and Janech,M. G. (2015a). Proteomic analysis of plasmafrom California sea lions (Zalophus Californianus) reveals apolipoprotein e as acandidate biomarker of chronic domoic acid toxicosis. PLoS ONE 10, e0123295.

Neely, B. A., Soper, J. L., Gulland, F. M. D., Bell, P. D., Kindy, M., Arthur, J. M.and Janech, M. G. (2015b). Proteomic analysis of cerebrospinal fluid in California

sea lions (Zalophus Californianus) with domoic acid toxicosis identifies proteinsassociated with neurodegeneration. Proteomics 15, 4051-4063.

Nesvizhskii, A. I. (2014). Proteogenomics: concepts, applications andcomputational strategies. Nat. Meth. 11, 1114-1125.

Pasing, Y., Colnoe, S. andHansen, T. (2017). Proteomics of hydrophobic samples:fast, robust and low-cost workflows for clinical approaches. Proteomics 17.

Paulo, J. A. (2013). Practical and efficient searching in proteomics: a cross enginecomparison. WebmedCentral 4, WMCPLS0052.

Pond, C.M. (2017). The evolution of mammalian adipose tissues. InAdipose TissueBiology (ed. M. E. Symonds), pp. 1-59. Cham: Springer International Publishing.

Russeth, K. P., Higgins, L. A. and Andrews, M. T. (2006). Identification of proteinsfrom non-model organisms using mass spectrometry: application to a hibernatingmammal. J. Proteome Res. 5, 829-839.

Sobolesky, P., Parry, C., Boxall, B., Wells, R., Venn-Watson, S. and Janech,M. G. (2016). Proteomic analysis of non-depleted serum proteins from bottlenosedolphins uncovers a high vanin-1 phenotype. Sci. Rep. 6, 33879.

Strandberg, U., Kakela, A., Lydersen, C., Kovacs, K. M., Grahl-Nielsen, O.,Hyvarinen, H. and Kakela, R. (2008). Stratification, composition, and function ofmarine mammal blubber: the ecology of fatty acids in marine mammals. Physiol.Biochem. Zool. 81, 473-485.

Yamaguchi, H., Hasegawa, K. and Esumi, M. (2013). Protein from the fractionremaining after RNA extraction is useful for proteomics but caremust be exercisedin its application. Exp. Mol. Pathol. 95, 46-50.

Van Dolah, F. M., Neely, M. G., McGeorge, L. E., Balmer, B. C., Ylitalo, G. M.,Zolman, E. S., Speakman, T., Sinclair, C., Kellar, N. M., Rosel, P. E. et al.(2015). Seasonal variation in the skin transcriptome of common bottlenosedolphins (Tursiops truncatus) from the northern Gulf of Mexico. PLoS ONE 10,e0130934.

Zhang, J., Xin, L., Shan, B., Chen, W., Xie, M., Yuen, D., Zhang, W., Zhang, Z.,Lajoie, G. A. and Ma, B. (2012). PEAKS DB: de novo sequencing assisteddatabase search for sensitive and accurate peptide identification. Mol. Cell.Proteomics 11, M111.010587.

Zhang, Y., Fonslow, B. R., Shan, B., Baek, M.-C. and Yates, J. R. (2013). Proteinanalysis by shotgun/bottom-up proteomics. Chem. Rev. 113, 2343-2394.

8

METHODS & TECHNIQUES Biology Open (2018) 7, bio036731. doi:10.1242/bio.036731

BiologyOpen

by guest on July 12, 2020http://bio.biologists.org/Downloaded from